University of Massachusetts Amherst ScholarWorks@UMass Amherst

Astronomy Department Faculty Publication Series Astronomy

2010 THE HUBBLE SPACE TELESCOPE WIDE FIELD CAMERA 3 EARLY RELEASE SCIENCE DATA: PANCHROMATIC FAINT OBJECT COUNTS FROM 0.2{2 MICRONS WAVELENGTH R Windhorst

S Cohen

N Hathi

P McCarthy JR

R Ryan

See next page for additional authors

Follow this and additional works at: https://scholarworks.umass.edu/astro_faculty_pubs Part of the Astrophysics and Astronomy Commons

Recommended Citation Windhorst, R; Cohen, S; Hathi, N; McCarthy, P JR; Ryan, R; Yan, H; Baldry, I; Driver, S; Frogel, J; Hill, D; Kelvin, L; Koekemoer, A; Mechtley, M; O'Connell, R; Robotham, A; Rutkowski, M; Seibert, M; Tuffs, R; Balick, B; Bond, H; Bushouse, H; and Calzetti, D, "THE HUBBLE SPACE TELESCOPE WIDE FIELD CAMERA 3 EARLY RELEASE SCIENCE DATA: PANCHROMATIC FAINT OBJECT COUNTS FROM 0.2{2 MICRONS WAVELENGTH" (2010). Astronomy Department Faculty Publication Series. 956. Retrieved from https://scholarworks.umass.edu/astro_faculty_pubs/956

This Article is brought to you for free and open access by the Astronomy at ScholarWorks@UMass Amherst. It has been accepted for inclusion in Astronomy Department Faculty Publication Series by an authorized administrator of ScholarWorks@UMass Amherst. For more information, please contact [email protected]. Authors R Windhorst, S Cohen, N Hathi, P McCarthy JR, R Ryan, H Yan, I Baldry, S Driver, J Frogel, D Hill, L Kelvin, A Koekemoer, M Mechtley, R O'Connell, A Robotham, M Rutkowski, M Seibert, R Tuffs, B Balick, H Bond, H Bushouse, and D Calzetti

This article is available at ScholarWorks@UMass Amherst: https://scholarworks.umass.edu/astro_faculty_pubs/956 Submitted to the Astrophysical Journal Supplement Series, May, 2010 Preprint typeset using LATEX style emulateapj v. 11/10/09

THE HUBBLE SPACE TELESCOPE WIDE FIELD CAMERA 3 EARLY RELEASE SCIENCE DATA: PANCHROMATIC FAINT OBJECT COUNTS FROM 0.2–2 MICRONS WAVELENGTH Rogier A. Windhorst 1, Seth H. Cohen 1, Nimish P. Hathi 2, Patrick J. McCarthy 3, Russell E. Ryan, Jr. 4, Haojing Yan 5, Ivan K. Baldry 6, Simon P. Driver 7, Jay A. Frogel 8, David T. Hill 7, Lee S. Kelvin 7, Anton M. Koekemoer 9, Matt Mechtley 1, Robert W. O’Connell 10, Aaron S. G. Robotham 7, Michael J. Rutkowski 1, Mark Seibert 3, Richard J. Tuffs 11, Bruce Balick 12, Howard E. Bond 9, Howard Bushouse 9, Daniela Calzetti 13, Mark Crockett 14, Michael J. Disney 15, Michael A. Dopita 16, Donald N. B. Hall 17, Jon A. Holtzman 18, Sugata Kaviraj 14, Randy A. Kimble 19, John W. MacKenty 9, Max Mutchler 9, Francesco Paresce20, Abihit Saha 21, Joseph I. Silk 14, John Trauger 22, Alistair R. Walker 23, Bradley C. Whitmore 9, & Erick Young 24 Submitted to the Astrophysical Journal Supplement Series, May, 2010

ABSTRACT We describe the Hubble Space Telescope (HST) Wide Field Camera 3 (WFC3) Early Re- lease Science (ERS) observations in the Great Observatories Origins Deep Survey (GOODS) South field. The new WFC3 ERS data provide calibrated, drizzled mosaics in the mid-UV filters F225W, F275W, and F336W, as well as in the near-IR filters F098W (Ys), F125W (J), and F160W (H) in 1–2 HST orbits per filter. Together with the existing HST Advanced Camera for Surveys (ACS) GOODS-South mosaics in the BVi’z’ filters, these panchromatic 10-band ERS data cover 40–50 square arcmin from from 0.2-1.7 µm in wavelength at 000.07– 00 00 < 0.15 FWHM resolution and 0.090 multidrizzled pixels to depths of AB ∼ 26.0–27.0 mag < (5-sigma) for point sources, and AB ∼ 25.5–26.5 mag for compact galaxies. In this paper, we describe: a) the scientific rationale, and the data taking plus reduction procedures of the panchromatic 10-band ERS mosaics; b) the procedure of generating object catalogs across the 10 different ERS filters, and the specific star-galaxy separation techniques used; and c) the reliability and completeness of the object catalogs from the WFC3 ERS mosaics. The excellent 000.07–000.15 FWHM resolution of HST/WFC3 and ACS makes star- < galaxy separation rather straightforward over a factor of 10 in wavelength to AB ∼ 25–26 mag from the UV to the near-IR, respectively. Our main scientific results are: 1) We present the resulting Galactic star counts and galaxy counts in 10 different filters. < From the ERS data, these could be accurately determined from AB'19 mag to AB ∼ 26 from the mid-UV to the near-IR, respectively. 2) Both the Galactic stars counts and the galaxy counts show mild but significant trends of decreasing count slopes from the mid–UV to the near-IR over a factor of 10 in wavelength. 3) We combine the 10-band ERS counts with the panchromatic GAMA counts at the bright < < end (10 ∼ AB ∼ 20 mag), and with the ultradeep BVizYJH HUDF counts and other avail- < < able HST UV counts at the faint end (24 ∼ AB ∼ 30 mag). The galaxy counts are now well < < measured over the entire magnitude range 10 ∼ AB ∼ 30 mag from 0.2–2 µm in wavelength. 4) We fit simple galaxy evolution models to these panchromatic galaxy counts over this entire flux range, using the best available 10-band local galaxy luminosity functions (LFs) from the GAMA survey, as well as simple prescriptions of luminosity and/or density evolution. < < While these models can explain each of the 10-band counts from 10 ∼ AB ∼ 30 mag well, no single one of these simple models can explain the counts over this entire flux range in all 10 filters simultaneously. arXiv:1005.2776v1 [astro-ph.CO] 16 May 2010 5) The 10-band panchromatic ERS data base is very rich in structural information at all rest-frame wavelengths where young or older stars shine during the peak epoch in the cosmic star-formation rate (z'1–2), and constitutes a unique new HST data base for the community to explore in the future. Subject headings: galaxies: evolution — galaxies: counts — galaxies: luminosity function, mass function — infrared: galaxies — ultraviolet: galaxies

[email protected] 1 School of Earth and Space Exploration, Arizona State California, Riverside, CA 92521 University, P.O. Box 871404, Tempe, AZ 85287 3 Observatories of the Carnegie Institution of Washing- 2 Department of Physics & Astronomy, University of ton, Pasadena, CA 91101-1292 2

4 Department of Physics, University of California, One Shields Avenue, Davis, CA 95616 5 Center for Cosmology and AstroParticle Physics, The Ohio State University, Columbus, OH 43210 6 Astrophysics Research Institute, Liverpool John Moores University, Birkenhead CH41 1LD, United King- dom 7 School of Physics and Astronomy, University of St An- drews, Fife KY16 9SS, UK 8 Association of Universities for Research in Astronomy, Washington, DC 20005 9 Space Telescope Science Institute, Baltimore, MD 21218 10 Department of Astronomy, University of Virginia, Charlottesville, VA 22904-4325 11 Max Planck Institute for Nuclear Physics (MPIK), Saupfercheckweg 1, D-69117 Heidelberg, Germany 12 Department of Astronomy, University of Washington, Seattle, WA 98195-1580 13 Department of Astronomy, University of Mas- sachusetts, Amherst, MA 01003 14 Department of Physics, University of Oxford, Oxford OX1 3PU, United Kingdom 15 School of Physics and Astronomy, Cardiff University, Cardiff CF24 3AA, United Kingdom 16 Research School of Astronomy & Astrophysics, The Australian National University, Weston Creek, ACT 2611, Australia 17 Institute for Astronomy, University of Hawaii, Hon- olulu, HI 96822 18 Department of Astronomy, New Mexico State Univer- sity, Las Cruces, NM 88003 19 NASA–Goddard Space Flight Center, Greenbelt, MD 20771 20 INAF–IASF Bologna, Via Gobetti 101, 40129 Bologna, Italy 21 National Optical Astronomy Observatories, Tucson, AZ 85726-6732 22 NASA–Jet Propulsion Laboratory, Pasadena, CA 91109 23 Cerro Tololo Inter-American Observatory, La Serena, Chile 24 NASA–Ames Research Center, Moffett Field, CA 94035 3

1. INTRODUCTION verse, amongst many of its possible science goals. 1.1. Scientific Background With the exquisite spectral coverage (0.2–1.7 µm), very high spatial resolution (000.04–000.16 FWHM), The study of the formation and evolution of galax- fine pixel sampling (000.039–000.13 FWHM), and high ies and large scale structure are amongst the most < sensitivity of WFC3 (AB ∼ 26–27 mag in 2 orbits; active interfaces between theory and observation in 10-σ for point sources), many new interesting ques- modern astrophysics. Galaxies are believed to have tions and outstanding problems can be addressed formed gradually over cosmic time from a combina- with the WFC3 ERS data. By sampling the vac- tion of merging and internal star-formation (Hop- uum UV with high sensitivity and the very high kins et al. 2006), regulated by feedback from stellar angular resolution afforded by the diffraction lim- winds, supernovae, and/or AGN (Scannapieco et al. ited 2.4 m Hubble Space Telescope, WFC3 can ob- 2005; di Matteo, Springel, & Hernquist 2005). The serve star-forming regions in galaxies over most of origin of the Hubble sequence is not yet understood, the Hubble time. The near-IR channel on WFC3 al- but is likely related to the balance between major lows one to do restframe visible band photometry of mergers versus minor accretion events and steady distant galaxies to very low stellar masses, and over infall (Conselice 2003; Hopkins et al. 2010) The crit- areas large enough to provide representative sam- ical epoch for the assembly of massive galaxies ap- ples. Together, the panchromatic images produced pears to be the ∼4 Gyr span from redshift z'3 to by WFC3 allows the user to decompose distant z'1 (e.g., Driver et al. 1998). > galaxies into their constituent substructures, exam- At redshifts ∼ 2–3, deep Hubble Space Telescope ine their internal stellar populations, and help con- (HST) imaging surveys and ground-based spec- strain their dust content. In this early release sci- troscopy have revealed a paucity of massive galax- ence program, the UVIS and IR channels of WFC3 ies and few classical disks or spheroids (Law, 2007, are used to provide a small, but representative sam- 2009). In contrast, large ground-based spectro- pling of the capabilities of WFC3 to examine the < scopic surveys targeting z ∼ 1 — coupled with HST formation and evolution of galaxies in the critical imaging — have shown that by this epoch massive galaxy assembly epoch of z'1–3, when the universe galaxies are largely mature, and that the Hubble se- was only 6–2 Gyrs old, respectively. quence is well developed (e.g., Abraham et al. 1995, The high spatial resolution panchromatic WFC3 2007; Driver et al. 1998; Lilly et al. 1998). While ERS data set allows the user to gauge the growth of there may be about a ∼50% growth in the stellar galaxies through star-formation and merging. High mass in spheroids in the last ∼7 Gyr, the process precision photometric and low-resolution spectro- of major galaxy assembly was largely completed by scopic grism redshifts allow one to make accurate z'1 (e.g., de Lucia et al. 2006; Dickinson et al.2003). determinations of the faint-end of the galaxy lumi- The interim period — from redshifts of z∼3 to z∼1 nosity and mass functions, and helps shed light on — is the era in which most of the stellar mass merging and tidal disruption of stellar and gaseous in galaxies is accumulated (Dickinson et al. 2003), disks. The WFC3 images also allow one to make de- and when galaxies acquire the characteristic struc- tailed studies of the internal structure of galaxies, tural and dynamical properties that define them to- and the distribution of young and old stellar popula- day. The HST Wide Field Camera 3 (WFC3) was tions. This program demonstrates the unique power designed and built to study this critical epoch of of WFC3, by applying its many diverse modes and galaxy assembly in detail. This has become one of full panchromatic capability to forefront problems the driving reasons of the ERS project. in astrophysics. Details of HST ERS program (PID #11359; PI R. O’Connell) can be found on this 1.2. Science Aspects of the WFC3 ERS program URL25. The current ERS program was specifically WFC3 was successfully installed into HST on May conceived to make maximum use of the WFC3 ca- 14, 2009, by the astronauts on-board Space Shut- pabilities — while the instrument was designed and tle Discovery during the Space Transportation Sys- constructed made from 1998–2008 — and to al- tem mission 125 (STS-125). This shuttle mission low for optimal comparison of the intermediate and was the fifth Servicing Mission of HST, for histor- high redshift galaxy samples identified in the cur- ical reasons referred to as SM4. Many of the cur- rent ERS program to nearby galaxy HST programs. rent co-authors were members of the WFC3 Sci- In §2 of this paper, we present the WFC3 ERS ence Oversight Committee (SOC) from July 1998 survey strategy, the ERS filters used and their through November 2009. Our main role as SOC achieved depths. In §3, we present the ERS obser- was to define the WFC3 science requirements and vations in both the WFC3 UVIS and IR channels, goals, monitor them during the pre-launch phases and the ERS pointings and their areal coverage. In of the project, and to oversee the design, implemen- §4, we present the WFC3 ERS data reduction pro- tation, integration, plus ground-based and on-orbit cedures and their reliability and current limitations. testing of the WFC3 instrument. In §5, we present the ERS object finding proce- The WFC3 was designed to provide a unique dures and catalog generation, and the star-galaxy opportunity to make the connection between the galaxy populations in the local and distant uni- 25 http://www.stsci.edu/cgi-bin/get-proposal-info 4 separation procedure used in the 10 ERS filters. In the redshift range z'1–3. The unique new data that §6, we present the panchromatic ERS star counts the WFC3 UVIS channel can provide are high reso- < and galaxy counts from 0.2–1.7 µm to AB ∼ 26-27 lution, very sensitive near-IR photometry over fields mag, and compare these to the 10-band ground- larger than those possible with HST NICMOS, or < < based GAMA counts from 10 ∼ AB ∼ 20 mag at the with adaptive optics from the ground (e.g., Stein- bright end, and the HUDF counts in BVizYJH fil- bring et al. 2004; Melbourne et al. 2005). < < ters from 24 ∼ AB ∼ 30 mag at the faint end. We also present the panchromatic ERS images of interesting 2.2. ERS Grisms Used individual objects. In §7, we summarize our main results and conclusions. Throughout this paper, we In addition to these broad-band filters, we used use WMAP-year7 cosmology (Komatsu et al. 2010), the WFC3 near-IR grisms G102 and G141 to ob- or H =71 km s−1 Mpc−1 ,Ω =0.26, and Λ=0.74. tain slitless spectroscopy of hundreds of faint galax- 0 o ies at a spectral resolution of R∼130–210. The 2. WFC3 AND ITS CAPABILITIES WFC3 near-IR grism data can trace the primary in- dicators of star-formation — the Lyman-α and H-α 2.1. The ERS Filter Set emission-lines — in principle over the redshift range In the current ERS program, the unique panchro- from z'5–13 and z'0.2–1.7, respectively. WFC3 can also trace the Lyman plus the rest-frame UV matic capabilities of WFC3 are used to survey the continuum slope, as well as the Balmer plus 4000 A-˚ structure and evolution of galaxies at the peak of breaks over the redshift range z'1–9 and z'0–2.5, the galaxy assembly epoch at z'1–3. Deep ul- respectively. The ERS grism program thus at least traviolet and near-IR imaging, and slitless near-IR covers the peak of the cosmic star formation history spectroscopy of existing deep multi-color GOOD- < < at redshifts 1 ∼ z ∼ 2, using some of the most impor- S/ACS fields are used to gauge star-formation and tant star-formation and post-starburst indicators, the growth of stellar mass as a function of galaxy while also providing some metallicity-independent morphology, structure and surrounding density in reddening indicators. Both IR grism dispersers pro- < < the critical cosmic epoch at redshifts 1 ∼ z ∼ 3. vide capabilities, that cannot be reproduced from The total HST filter set provided by the WFC3 the ground: slitless spectroscopy of very faint ob- ERS imaging of the GOODS fields is shown in < jects (AB ∼ 25–26 mag) over a contiguous wide spec- Fig. 1a. WFC3 adds the F225W, F275W and tral range in the near-IR, that is not affected by F336W filters in the WFC3 UVIS channel, and the night-sky lines. F098M, F125W and F160W filters in the WFC3 IR The addition of the low-resolution (R ' 130–210) channel. Together with the existing GOODS ACS near-IR spectra also allows the user to obtain ac- F435W, F606W, F775W and F850LP images, the curate spectro-photometric redshifts for objects as new WFC3 UVIS and IR filters provide a total of 10 faint as AB'25 mag (cf. Pirzkal et al. 2004; Mal- HST filters that span the wavelength range λ'0.2– hotra et al. 2005, Straughn et al. 2008, 2009, Co- 1.7 µm nearly contiguously. We refer to this entire hen et al. 2010). The redshift precision is expected survey — including the new WFC3 data and the < to be a few percent or σz/(1+z) ∼ 0.04 (Ryan et al. existing GOODS-S ACS v2.0 data in BViz — here- 2007), allowing accurate determination of the faint- after as the “ERS”. Details of the GOODS survey end of the galaxy luminosity function (LF), and can be found in Giavalisco et al. (2004) and refer- to find potential fainter members of known small- ences therein. Fig. 1a compares the ERS filters scale galaxy clustering and large-scale structures in to the spectral energy distribution of model galax- the redshift range z'0.5–3 across the WFC3 ERS ies at z'1.0, 1.5 and 2.5 with ages of 3, 2 and 1 mosaics. Such grism data cannot provide precise Gyr, respectively. Fig. 1a illustrates that the WFC3 spectroscopic redshift measurements to identify true UV filters are particularly powerful in accessing the large scale structure, but it allows one to do sta- Lyman-α forest and Lyman continuum breaks at tistical studies of spatially close objects with simi- redshifts z'1–3. lar spectro-photometric redshifts, that have a high The ERS images in the WFC3 UVIS filters likelihood to be at similar distances, which e.g., F225W, F275W, and F336W are used to identify < helps determine the epoch-dependent galaxy major galaxies at z ∼ 1.5 from their UV continuum breaks, merger rate (Ryan et al. 2008). Details of the ERS and provide star-formation indicators tied directly > grism data reduction and its analysis are given by to both local and z ∼ 3 galaxy populations, which are Straughn et al. (2010) and Kuntschner et al. (2010). the ones best observed through their Lyman breaks > WFC3 UVIS G280 UV-prism observations were from the ground at λ ∼ 350 nm. The critical new not done as part of the ERS, because of its much data that the WFC3 UVIS channel can provide are lower throughput and the significant overlap of its thus very high resolution, deep vacuum mid-UV im- < < spectral orders (Bond & Kim Quijano 2007; Wong ages from 0.2 ∼ λ ∼ 0.3 µm. et al. 2010). The ERS images in the WFC3 near-IR filters F098M, F125W and F160W are used to probe the Balmer and 4000 A˚ breaks and stellar mass function 2.3. WFC3 Detectors and Achieved ERS 9 Sensitivities well below 10 M for mass-complete samples in 5

The very sensitive ERS observations were made panel). This beneficial, but significant discrepancy possible by assuring that the WFC3 had the best is believed to result from uncertainties in the ab- possible CCD and near-IR detectors. WFC3s UV— solute calibration procedure of the optical stimu- blue optimized CCDs were chosen specifically to lus used in the Thermal Vacuum tests of WFC3. complement those of ACS. They were made by E2V The cause of this 6–18% discrepancy is currently in the UK, and are thinned, backside illuminated, being investigated, and lessons learned will be ap- CCD detectors with 2k×4k 15 µm (000.0395) pixels, plied to the upcoming ground-based calibrations of covering the wavelength range 200–1000 nm with the James Webb Space Telescope Thermal Vacuum > QE ∼ 50% throughout (Wong et al. 2010). absolute throughput measurements. The WFC3 near-IR detectors were Teledyne HgCdTe infrared detectors, and were MBE-grown, 3. THE ERS DATA COLLECTION STRATEGY substrate removed, with Si CMOS Hawaii-1R mul- tiplexers, and have 1k×1k 18 µm (000.130) pixels, 3.1. ERS Pointings and Areal Coverage covering the wavelength range 800–1730 nm with To make the best use of the limited HST space- > QE ∼ 77% throughout (Wong et al. 2010). Fur- craft time, the WFC3 ERS program is most effi- ther specifications of the WFC3 detectors are listed ciently carried out in existing HST Treasury fields. where relevant below. The two Great Observatories Origin Deep Sur- Table 1 summarizes the resulting excellent WFC3 vey (GOODS) fields are among the best candidate sensitivities from our relatively short ERS expo- fields. They contain deep four-color ACS imaging sures. The table lists the number of orbits per filter (Giavalisco et al. 2004; Dickinson et al. 2004), and and the 5-σ depths in AB magnitudes and Fν units. ACS slitless G800L grism spectra covering wave- A net exposure time of 2600–2700 seconds was avail- lengths from ∼0.55–0.95 µm (cf. Pirzkal et al. able in each HST orbit for on-source ERS observa- 2004; Malhotra et al. 2005, Straughn et al. 2008, tions. In Fig. 1b, the equivalent depths are plotted 2009). There is also a wealth of ground- and space- in physical terms, by comparing with with spectral based data, such as deep VLT K-band imaging synthesis models of Bruzual & Charlot (2003) at the (Labb´eet al. 2003), a large number of VLT spec- three fiducial redshifts, following Ryan et al. (2007, tra (Vanzella et al. 2006), deep Chandra X-ray im- 2010). Simple stellar populations models are plot- ages (Giacconi et al. 2002), Spitzer photometry with ted with single bursts, or exponentially declining IRAC and MIPS (Dickinson et al. 2006; Yan et al. star-formation rates with an e-folding time of 1 Gyr. 2004, 2005), and deep VLA and ATCA radio images Fig. 2b shows the predicted spectral energy distri- (cf. Afonso et al. 2006, Kellermann et al. 2008). butions for models with ages ranging from 10 Myr to 3 Gyr, along with the 5-σ depths of the WFC3 3.2. ERS Mosaics and WFC3 Scheduling ERS program. The three panels represent redshifts z=1.0, 1.5 and 2.0, and models with stellar masses of Given the constraint on the total amount of time 9 9 10 available — and the desire to use the above WFC3 M=10 , 4×10 , and 10 M , respectively. These SED tracks illustrate the intended SED and mass panchromatic filter set — in the allotted HST time sensitivity of the WFC3 ERS observations as a func- of 104 orbits, the ERS program could survey one 4×2 WFC3 mosaic covering 100 ×50 or roughly 50 tion of cosmic epoch. Galaxies with ongoing star- 2 formation, even with fairly large ages, are easily de- arcmin to 5-sigma depths of mAB ' 26.0 mag in the three bluest wide-band UVIS filters, and one tected in the WFC3 mid-UV observations. At z'2, 0 0 ∗ 5×2 WFC3 mosaic covering 10 ×4 or roughly 40 a maximally old τ'1 model with a mass of ∼0.3 M 2 is detectable above the WFC3 detection threshold arcmin to 5-sigma depths of mAB ' 27.0 mag in in the F336W filter, while at z'1, the WFC3 ERS the three near-IR filters. These angular scales probe comoving scales of roughly 5–10 Mpc, and provides can detect young star forming galaxies with masses < 7 ∗ a sample of 2000–7000 galaxies to AB ∼ 26–27 mag as low as a few×10 M , or about a M∼0.01 M galaxy in that filter. in the panchromatic ERS images. The IR images Table 2 summarizes the high level parameters of were dithered to maximally match the UVIS field the HST instrument modes, the ERS filters used, of view. The layout of these ERS UVIS and IR the filter central wavelength λ, its width ∆λ or mosaics is given in Fig. 3, compared to the other FWHM — both in µm — the PSF FWHM as a ACS data that are available in the GOODS-South function of wavelength, the zero-points for all 10 field, as well as the ACS images taken in parallel to ERS filters in units of 1.0e− /sec, as well as the zo- the ERS data. diacal sky-background measured in each ERS filter. Fig. 3 shows the ACS z’-band (F850LP) mosaic The GOODS BViz sky-background values are from of the entire GOODS-South field, and the outline Hathi et al. (2008). of the acquired WFC3 pointings, as well as the lo- Fig. 2 shows that on average, the on-orbit cations of some of the other ACS data sets. The WFC3 UVIS sensitivity is 6–18% higher than the WFC3 ERS mosaic pointings cover the Northern predicted pre-launch sensitivity from the ground- ∼30% of the GOODS-South field (Fig. 3). The 8 based thermal vacuum test (left panel), and the on- ERS pointings are contiguous, with a tiling that can orbit WFC3 UVIS sensitivity is 9–18% higher (right be easily extended to the South in future WFC3 GO programs. 6

HST scheduling required that this ERS program The measured on-orbit UVIS read-noise levels are be split into several visits. Simple raster patterns 3.1e− in Amp A, 3.2e− in Amp B, 3.1e− in Amp were be used to fill out the WFC3 IR mosaic, and C, and 3.2 e− in Amp D, respectively, or on aver- to improve sky plus residual dark-current removal. age about 3.15 e− per pixel across the entire UVIS Only mild constraints were applied to the original array. HST roll angle (ORIENTs) to maximize overlap be- (2) A null dark frame was applied, since the only tween the northern part of the GOODS-South field, available darks at the time were from the ground- and to allow HST scheduling in the permitted ob- based thermal vacuum testing in 2007–2008, and serving interval for the ERS (mid-Sept.–mid-Oct. were not found to reduce the noise in the output 2009). The IR and UVIS images were constrained ERS frames. Hence, the subtraction of actual on- to have the same ORIENT, to ensure that a uniform orbit 2D dark-frames was omitted until better, high WFC3 mosaic could be produced at all wavelengths. signal-to-noise ratio on-orbit dark-frames have been The slight misalignment of the Northern edge of the accumulated in Cycle 17 and beyond. Instead, a existing ACS GOODS-South mosaics and the new constant dark-level of 1.5 e− /pix/hr was subtracted WFC3 ERS mosaics in Fig. 3 was due to the fact from all the images, as measured from the average that the HST ORIENT constraints had changed in dark-current level in the few on-orbit dark-frames the late summer of 2009, because the ERS obser- available thus far. This dark-level is about 5× vations were postponed by several weeks due to a higher than the ground-based thermal vacuum tests change in the HST scheduling constraints. had suggested, but still quite low enough to not add The area of overlap between the individual WFC significant image noise in an average 1300 sec UVIS ERS mosaic pointings is too small to identify tran- exposure. sient objects (e.g. SNe and variable AGN), since only about 1–2% of all faint field objects show point- (3) A bad pixel file (tb41828mi bpx.fits) was cre- source variability (Cohen et al. 2006). However, it ated (by H. Bushouse) and updated over the one is useful in the subsequent analysis to verify the available in OPUS at the time, and applied to all positions of objects, and so verify the instrument the UVIS images. geometrical distortion corrections (GDCs) used, as (4) All flat-fields came from the 2007–2008 WFC3 discussed in §4 and Appendix A. thermal vacuum ground tests, and had high signal- to-noise ratio, but left some large-scale gradients in 4. WFC3 UVIS AND IR DATA PROCESSING the flat-fielded data, due to the illumination differ- The WFC3 ERS data processing was carried out ence between the thermal vacuum optical stimulus with the STScI pipeline calwf3. The WFC3 ERS and the real on-orbit WFC3 illumination by the zo- data set also provided tests of the STSDAS pipeline diacal sky. For each passband, the mean UV-sky under realistic conditions. This process was started was removed from the individual ERS images, and well before the SM4 launch in the summer of 2008 the resulting images were combined into a median with pre-flight WFC3 thermal vacuum calibration image in each UV filter. The large scale gradients data, and continued through the late summer and from this illumination difference correspond to a fall of 2009, when the first ERS data arrived. The level of about ∼5–10% of the on-orbit zodiacal sky raw WFC3 data was made public immediately, and background, and have very low spatial frequency. the OTF pipe-lined and drizzled WFC3 mosaics will This situation will be remedied with on-orbital in- be made public via MAST at STScI when the fi- ternal flats and sky-flats, that will be accumulated nal flight calibrations — as detailed below — have during the course of Cycle 17 and beyond. Since been applied. The specific pipeline corrections that the UV sky is very low to begin with (∼25.46 mag were applied to the WFC3 ERS images are detailed arcsec−2 in F225W, 25.64 mag arcsec−2 in F275W, here. Unless otherwise noted below, the latest ref- and 24.82 mag arcsec−2 in F336; see Table 2 here < erence files from the WFC3 Calibration web-page and also Windhorst et al. 2002), these residual ∼ 5– were used in all cases, and are on this URL26 10% sky-gradients affect the object photometry only > at the level of AB ∼ 27–28 mag, i.e., well below 4.1. The Main WFC3 Pipeline Corrections the catalog completeness limits discussed in §4, and 4.1.1. WFC3 UVIS Pipeline Corrections hence does not significantly impact the construc- tion of complete samples from our UVIS images to All raw WFC3 UVIS data were run through the < AB ∼ 25.5–26.0 mag. standard STSDAS calwf3 pipeline as following: We suspect, but have at this stage not been able (1) The current best available WFC3 UVIS super- to prove with the currently available data, that this bias (090611120 bia.fits) was subtracted from all remaining low-level sky gradient is of multiplicative images. This is an on-orbit super-bias created from and not of additive nature. Once we have been 120 UVIS bias frames, each of which was unbinned, able to demonstrate this with the full suite of Cy- and used all four on-chip amplifiers (see URL27. cle 17 sky-flats, we will re-process all the UVIS data again, and remove these low-level gradients 26 www.stsci.edu/hst/observatory/cdbs/SIfileInfo/WFC3/reftablequeryindexaccordingly. For now, they are not visible in the 27 http://www.stsci.edu/hst/wfc3/new uvis bias lbn high quality, high contrast color reproductions of 7

Figs. 4–5. Hence, they do not significantly affect F225W, F275W and F336W filters, the higher S/N- the subsequent object-finding and their surround- ratio GOODS ACS B-band images were used as the < ing sky-subtraction procedures (to AB ∼ 25.5–26.0 detection image. This also provided an astromet- mag), which assume linear remaining sky-gradients. ric reference frame that was matched as closely as This is corroborated by the quality of the panchro- possible to the wavelengths of the UVIS filters used matic object counts discussed below, and its con- in these observations. Because the GOODS B-band > sistency with the counts from other authors at the images reach AB ∼ 28.0 mag and so go much deeper flux range of overlap. In other words, any remain- than the ERS UVIS images, they help optimally lo- ing low-level sky gradients do not significantly affect cate the objects in the ERS UVIS mosaics. the object catalogs generated for the current science Furthermore, the orbital F225W and F275W ERS < purposes to AB ∼ 25.5–26.0 mag. observations were designed to minimize possible Earth limb contamination. To guarantee the low- 4.1.2. WFC3 IR Pipeline Corrections est possible UV sky in the WFC3 images, only one exposure was obtained in F225W at the end of each The reduction of the ERS WFC3 IR data largely orbit, in contrast with the common practice of ob- followed the procedures as described in Yan et al. serving all exposures in the same filter in rapid suc- (2009). We used the calwf3 task included in the cession in subsequent orbits. (All three F336W ex- STSDAS package to process the raw WFC3 IR im- posures were obtained during the same single orbit ages, using the latest reference files indicated by the for a given ERS pointing). This manner of schedul- relevant FITS header keywords: ing indeed minimized the on-orbit UV sky away (1) We removed the artificial quadrature “pedestal” from the Earth’s limb (see Table 2), but somewhat signature (caused by the IR gain being corrected complicated the sub-sequent multidrizzle procedure, twice in the current calwf3 pipeline version), using since no “same-orbit” cosmic ray rejection could be the average gain value — < g >=1.54 over that of applied in F225W and F275W, to allow one to find the four amplifiers, which ranges from gi=1.52–1.56 a first slate of bright objects for image alignment. — as given in the WFC3 STAN (September 2009 As a result, the F225W and F275W ERS expo- issue; see URL28; see also Wong et al. 2010). Specif- sures taken in successive orbits needed to be aligned ically, after correcting for the fact that g=1.54 was with each other individually, in addition to their applied twice, the incremental multiplicative gain overall alignment onto the GOODS reference frame. values applied (once) were 1.004 for Quadrant 1 (up- For each ERS filter, the relative alignment between per left), g = 0.992 for Quadrant 2 (lower left) = exposures was achieved iteratively, starting with an 0.992, g=1.017 for Quadrant 3 (lower right), and initial partial run of multidrizzle (Koekemoer et al. g=0.987 for Quadrant 4 (upper right quadrant). 2002) to place each exposure onto a rectified pixel (2) For each passband, the mean near-IR sky was grid. These images were then cross-correlated with removed from the individual ERS images, and the each other, after median-filtering a smoothed ver- resulting images were combined into a median image sion of each UVIS exposure to reduce the impact in each near-IR filter. of cosmic rays and to identify the brighter real ob- jects. This ensured a relative alignment between (3) A smooth background gradient still persisted in the sequential orbital ERS exposures to the sub- the median image. This gradient was fitted by a pixel level, correcting offsets that were introduced 5-order Spline-3 function, and was then subtracted by the guide-star acquisitions and re-acquisitions at from the individual near-IR images. This sky gradi- the start of each successive orbit. ent is of order 1–2% of the zodiacal sky-background. Since the near-IR zodiacal sky background is about −2 4.2.2. The WFC3 UVIS Multidrizzle Procedure 22.61, 22.53, and 22.30 mag arcsec in YsJH, re- spectively. (see Table 2), these remaining WFC3 IR These first-pass aligned images were then run gradients do not affect the large-scale object find- through a full combination with multidrizzle ing and catalog generation at levels brighter than (Koekemoer 2002), which produced a mask of cos- AB'27.0 mag. mic rays for each exposure, together with a cleaned image of the field. The cosmic ray mask was used 4.2. WFC3 Astrometry and Multidrizzle to create a cleaner version of each exposure, by Procedures substituting pixels from the clean, combined im- 4.2.1. WFC3 UVIS Astrometry age. These were then re-run through the cross- correlation routine to refine the relative shifts be- The calibrated, flat-fielded WFC3 UVIS expo- tween the exposures, achieving an ultimate relative < sures were aligned to achieve astrometric registra- alignment between exposures of ∼ 2–5 mas. This tion with the existing GOODS ACS reference frame process was limited primarily by the on-orbit cosmic (Giavalisco et al. 2004; GOODS v2.0 from Gro- ray density, and the available flux in the faint UV gin et al. 2009). To generate accurate SExtrac- objects visible in each individual UVIS exposure. tor (Bertin & Arnouts 1996) object catalogs in the In the end, about 3 independent input pixels from 4 exposures contributed to one multidrizzle UVIS 28 http://www.stsci.edu/hst/wfc3/STAN 09 2009 < output pixel. Of these, typically ∼ 1–2% were re- 8 jected in the cosmic ray rejection, leaving typically formation populated in the image headers. This was 3 independent UVIS measurements contributing to done by using the multidrizzle software (Koekemoer one multidrizzle output pixel in both the F225W et al. 2002) distributed in the STSDAS.DITHER and F275W filters. In F336W, about 2.3 indepen- package. Similar to the processing of the UVIS im- dent input pixels from the 3 F336W exposures con- ages in §4.2.1, the GOODS version 2.0 ACS mosaics tributed to one output pixel during the multidrizzle were used as the astrometric reference. The only process. difference is that the GOODS ACS mosaics were After this relative alignment between exposures straightly 3 × 3 rebinned for comparison with the was successfully achieved, each set of exposures ERS IR data, giving a spatial resolution of 000.090 needed to be aligned to the absolute GOODS as- per pixel for all ERS images. trometric reference frame. This was achieved by As usual, the projected ERS IR images show non- generating catalogs from the cleaned, combined im- negligible positional offsets, which is mainly caused ages for each of the three ERS UVIS filters F225W, by the intrinsic astrometric inaccuracies of the guide F275W, and F336W, and matching them to the stars used in the different HST visits. Following Yan GOODS B-band catalog (Giavalisco et al. 2004; et al. (2009), about 6–12 common objects were were Grogin et al. 2009). This was done by solving for manually identified in each ERS IR input image linear terms (shifts and rotations) using typically and in the reference ACS z850 image, and we sub- ∼30–50 objects matched in each pointing, depend- sequently solved for X-Y shift, rotation, and plate ing on the UVIS filter used. This procedure suc- scale between the two. These transformations were cessfully removed the mean shift and rotational off- then input to multidrizzle, and the drizzling process sets for each visit relative to the GOODS astro- was re-run to put each input image onto the pre- metric frame. Multidrizzle also produced “rms-” or specified grid. We set the drizzling scale (“pixfrac”) “weight”-maps (Koekemoer 2003), which are essen- to 0.8, so that in the IR, about 5 input pixels from tial for the subsequent object detection (§5.1), and 6 exposures, or 20 independent measurements con- for the computation of the effective area, which is tributed to one multidrizzle output pixel. Of these, < needed for the object counts (see the figures in §6). typically ∼ 10% or 2 pixels were rejected in the cos- In order to perform matched aperture photome- mic ray rejection, leaving typically 18 independent try (see §5.2), our approach was to create images measurements contributing to one multidrizzle out- at all wavelengths at the same pixel-scale. Since put pixel. the IR data was necessarily created at 000.090 per pixel (see Appendix A), we created UVIS mosaics 4.3. Resulting ERS Mosaics and their Properties at that same pixel size. This essentially “smoothed” 4.3.1. The panchromatic 10-band ERS mosaics over the remaining issue of the geometrical distor- tion solution (see Appendix A), and created a suf- Fig. 5a. shows the panchromatic 10-band color ficient data product for the purpose of producing image of the entire ERS mosaic in the CDF-South. matched aperture photometric catalogs, reliable to- All RGB color images of the 10-band ERS data were tal magnitudes in all 10 ERS filters, performing made as following. First, the mosaics in all 10 filters z'1–3 dropout searches, and many other “total have the common WCS of the ACS GOODS v2.0 magnitude applications”. The performance of the reference frame to well within one pixel. Second, panchromatic ERS images for photometric redshift all images were rescaled to Fν units of Jy per pixel estimates is described by Cohen et al. (2010). Fur- using the AB zero-points of Table 2. Next, the blue ther details on the remaining uncertainties from the gun of the RGB images contained a weighted ver- UV geometrical distortion and its corrections are sion of the UVIS images in the F225W, F275W, and given in Appendix A and Fig. 4a–4b. F336W filters and the ACS F435W filter. The green The current 000.090 per pixel UVIS image mosaic gun contained a weighted version of the ACS F606W is referred to as “ERS version v0.7”. In the fu- and F775W images, which had the highest S/N- ture, when the UV geometrical distortion correc- ratio of all available images. The red gun contained tion is well measured and more dither points for a weighted version of the ACS z-band (F850LP) and the ERS UV images are available, we will make WFC3 IR F098M, F125W, and F160W images. All higher resolution images (000.030 per pixel) for ap- weighting was done with the typical image sky S/N- plications such as high-resolution faint-galaxy mor- ratio, sometimes adjusted so as to not overempha- phology, structure, half-light radii, and other high- size the deepest multi-epoch GOODS v2.0 images precision small-scale measurements of faint galaxies, in the V and i-filters. This procedure thus also re- and make these end-products available to the com- balanced the different sensitivities per unit time in munity. these filters, as shown in Fig. 1a–1b, and corrected for the fact that some filters overlap somewhat more than at the usual FWHM wavelength values (see 4.2.3. IR Astrometry and Multidrizzle Procedure Table 1). This is especially noticeable for the ACS The WFC3 IR images processed in section 3.1 z-band filter F850LP — which at the long wave- were first corrected for the instrument geometric length side is cut-off by the sharp decline in the QE- distortion, and then projected to a pre-specified as- curve of silicon — and the IR F098M filter, which trometric reference grid according to the WCS in- doesn’t have this problem at its blue side, but over- 9 laps with F850LP for about 40% of its OTA×QE×T Table 1 summarizes the exposure times and the ac- integral. (When the QE of the HgCdTe detectors tual achieved depth in each of the observed ERS produced by Teledyne increased from ∼10–20% in filters, while Table 2 also lists the effective area cov- > 2001 to ∼ 80% after 2005, the F098M filter became ered in each filter mosaic at the quoted depth. thus almost a replacement of the ACS z-band. The histograms of Fig. 8a give the cumulative In Fig 5a–5b, we used log(log) stretch to opti- distribution of the maximum pixel area that pos- mally display the enormous dynamic range of the sesses a specified fraction of total orbital exposure full resolution ERS color TIFF images. Fig. 5a time. These effective areas must be quantified in only displays the part of the common cross sections order to properly do the object counts in Fig. 11. between the 4×2 ERS UVIS mosaics, the GOODS After Multi-drizzling the ERS mosaics in the UV v2.0 ACS BViz, and the 5×2 ERS IR mosaics. Each and IR, these effective areas were computed from of the ERS mosaics are 8079 × 5540 pixels in total, the weight maps, which include the total exposure but only about 6500 × 3000 pixels or 9.75 × 4.5 or time, and the effects from CR-rejection, dithering, 43.875 arcmin2 is in common between the UVIS and and drizzling. Fig 8a shows that about 50 arcmin2 IR mosaics, and shown in Fig. 5a–5b. The area of of the UV mosaics has ∼80% of the UVIS exposure the individual UVIS mosaics that is used in each of time or ∼90% of the intended UV sensitivity. For the UV-optical galaxy counts of §6 is substantially reference, one 00.09 pixel could be composed of 5.2 larger than this, but the total usable area of the IR native WFC3 UVIS pixels times the number of ex- mosaics is comparable to the area shows in Fig. 5a. posures on that portion of sky. Due to overlapping Fig. 5b shows a zoom of the 10-band ERS color dithers (see Fig. 3), some pixels have more than the image in the CDF-South, illustrating the high res- total orbital exposure time contributing to their flux olution available over a factor of ten in wavelength, measurement. The histograms of Fig. 8b show the the very large dynamic range in color, and the sig- same as Fig. 8a, but for the six GOODS v2.0 mo- nificant sensitivity of these few orbit panchromatic saic tiles in BViz that overlap with the ERS. Fig. 8c ERS images. Further noteworthy objects in the shows the same as Fig. 8a, but for the ERS mosaics ERS images are discussed in §6.5 below. in the IR. About 40 arcmin2 has ∼80% of the expo- sure time in the ERS IR mosaics, or ∼90% of the in- 4.3.2. The panchromatic ERS PSFs tended IR sensitivity. The overall WFC3 UVIS – IR Fig. 6a shows a full color reproduction of a stellar sensitivity is 9–18% better than predicted from the image in the 10-band ERS color images in the CDF- ground-based thermal vacuum tests (see §2.1 and South, and Fig. 6b shows a double star. These im- Fig. 2), and so in essence 100% of the intended ex- ages give a qualitative impression of the significant posure time was achieved over the 50 arcmin2 UVIS dynamic range in both intensity and wavelength images and the 40 arcmin2 IR images. that is present in the ERS images. Fig. 7a show images in all 10 ERS filters of an isolated bright 5. CATALOG GENERATION FROM THE star that was unsaturated in all filters, and Fig. 10-BAND ERS MOSAICS 7b shows its 10-band stellar light-profiles. Table 2 5.1. Object Finding and Detection shows the stellar PSF FWHM values in the 10 ERS All initial catalogs were generated using SExtrac- bands. These include the contribution form the tor version 2.5.0 (Bertin & Arnouts 1996) in sin- OTA and its wavefront errors, the specific instru- gle image mode. It was necessary to change the ment pixel sampling or Modulation Transfer Func- parameters in SExtractor to handle the UVIS im- tion (MTF). ages slightly differently than the ACS/WFC and Table 2 and Fig. 7b show the progression of the WFC3/IR ones. This is due to several factors, HST λ/D values with wavelength in the 10 ERS both cosmetic and physical. The major difference is filters. Table 2 implies that the larger pixel values that galaxies appear much smoother in the IR and used in the multidrizzling of §4.1.2 indeed add to the clumpy in the UV (see Windhorst et al. 2002 for effective PSF diameter. Table 2 also shows that that a discussion of this effect for all Hubble types), so HST is diffraction limited in V-band and longwards, that star-forming regions in well-resolved galaxies while shortward of V-band, the PSF FWHM starts can be deblended into separate objects, if the de- to increase again due to mirror micro-roughness in blending is overly aggressive. Fortunately, most of the ultraviolet. At wavelengths shorter than the V- these intermediate redshift galaxies are not bright band, HST is no longer diffraction limited, resulting in the observed UV, causing the UVIS fields to ap- in wider image-wings, and a somewhat larger frac- pear rather sparse, and making deblending a minor tion of the stellar flux visible outside the PSF-core. > issue. The “red halo” at λ ∼ 0.8 µm is due to noticeable The UVIS images have a less uniform background Airy rings in the stellar images in the WFC3 IR (∼10–15% of sky), which as discussed in §4.1.1 can channel, and the well known red halo in the ACS z- possibly be improved when on-orbit sky-flat calibra- band (Maybhate et al. 2010) has an additional com- tion files become available. For these reasons, we ponent due to light scattered off the CCD substrate. adopted a seemingly low threshold for object detec- 4.3.3. The 10-band ERS Area and Depth tion with SExtractor, requiring 4 connected pixels that are 0.75σ above the background in the UV. 10

Due to the clumpy nature of the ERS objects in the noise. UV, we convolved the UVIS image with a Gaussian kernel of FWHM of 6.0 pixels for the object detec- 5.3. ERS Sample Completeness tion phase only. The deblending parameter, DE- Fig. 9b shows the completeness functions of the BLEND MINCONT, was set to 0.1, to assure that 10-band ERS images. These were derived by Monte real objects were not over-deblended. It should be Carlo insertion of faint point sources into the ERS emphasized that object crowding in not an issue in images, and plotting the object recovery ratio as these medium depth UV images, so that the use of a function of total magnitude. In summary, the a larger SExtractor convolution filter, a lower SB- WFC3 UVIS images are ∼50% complete for point- detection threshold, and less object deblending is > like objects with ∼ 5-σ detections in total flux for fully justified. < < AB ∼ 26.3 mag in F225W, AB ∼ 26.4 mag in F275W, The SExtractor input parameters for the ACS im- < and AB ∼ 26.1 mag in F336W. The WFC3 IR images ages and WFC3/IR images were the same. The de- < < are complete to AB ∼ 27.2 mag in F098M, AB ∼ 27.5 tection threshold was set to 1.5σ, again requiring 4 < mag in F125W, and AB ∼ 27.2 in F160W, as listed connected pixels above the threshold for catalog in- in Table 1. The actual star counts in §6 appear clusion. The convolution filter was a Gaussian with indeed complete to roughly these limits in the UV, FWHM of 3.0 pixels, and DEBLEND MINCONT but to slightly brighter limits in the IR due to the was set to 0.06. For all ten bands, the appropriate significant object confusion of red stars with faint RMS maps (see §4.2.2) were used in order to cor- red background galaxies. rectly account for image borders (Fig. 3), as well The actual UV galaxy counts in §6.2 turn over as to properly characterize the photometric uncer- at a level rather close to the above UVIS point- tainties and the effective areas for each mosaic (Fig. source sensitivity limits. This is due to the in gen- 8). eral more compact nature of the ERS objects and the extremely faint zodi sky-background in the UV. 5.2. Object Extraction In the IR, the galaxy counts turn over at levels a Two post-processing steps were taken to clean the little brighter than the 5-σ point-source sensitiv- catalogs of residual artifacts. First, due to the rela- ity limits listed in Table 2. Despite the generally tively small number of exposures per filter, there rather compact nature of the faint red field galaxies were residual cosmic rays at the borders of the selected in YsJH, this is due to the higher zodiacal dither patterns, and in the chip gaps. These were sky background values in the near-IR (see Table 2). cleaned by masking out all objects in regions where 5.4. Star Galaxy separation in the ERS Mosaics the number of exposures was less than three. There- fore the total area coverage in each filter (Table 2 Stars and galaxies were separated as following. In & Fig. 8) is slightly different. Secondly, SExtrac- each filter, the stellar locus was defined as shown in tor will detect the diffraction spikes of bright point the usual plot of object FWHM versus total AB- sources. These are removed by searching for bright magnitude. Fig. 10 illustrates our star-galaxy sep- objects with FWHM near the size of the PSF, and aration procedure used for the 10-band ERS im- searching for surrounding objects in the catalog that ages. Objects with re Petrosian factor of 2.0, and Kron (1980) magnitudes and they have re ∼ (1+)×FWHM, where “1+” is were measured using a Kron-factor of 2.5 (SExtrac- a weak function of total S/N-ratio, as shown by the tor parameter MAG AUTO). For the galaxy counts solid, nearly vertical slanted lines in Fig. 10. in §6.2, Kron magnitudes were used, because the Next, all ten independent ERS object catalogs correction to total magnitude — preferably using a — each with their independent star-galaxy separa- Sersic extrapolation — is less than that typically tion and photometry — were merged by astrometric required for a Petrosian magnitude. Within the cross-matching. In order to maximize the 10-band errors, the Kron and Sersic based number counts information, we considered an object to be a star are indistinguishable (Hill et al. 2010b). The total if it was classified as stellar in at least three filters. magnitude errors in the ERS as a function of total This provides robust stellar classifications, perhaps AB-magnitude were determined from the SExtrac- to fainter limits than can be done in single filter tor errors in each filter, which are a combination of alone. multidrizzle weight-map errors and the image shot- From the resulting ERS objects counts in Fig. 11

11a, we can see that the stars separate out well break. The values of the best fit power-law star < from the galaxies to about AB ∼ 25.5–26.0 mag, and count slope is in general of order 0.03–0.05 dex/mag, somewhat fainter in the ACS BViz filters and the i.e. well below Euclidean value of 0.6 dex/mag. WFC IR filters. To these flux levels, the star counts This shows that to a depth of AB'25–26 mag at follow a power-law rather well, as shown in Fig. 11a. all wavelengths in the range 0.2–2 µm, the ERS im- The choice of three filters to require an object to be ages are quickly looking through most of the thick classified as a star — as opposed to all ten filters disk of our Galaxy at these latitudes (bII '–54◦) in — was made because stellar SEDs do not span a the direction of the Galactic anti-center (lII =224◦; large range in wavelength, and because we are merg- Ryan et al. 2005; Beckwith et al. 2006). This trend ing the data from three different detectors (WFC3 was also seen in the red ACS imaging parallels of UVIS CCD, ACS/WFC CCD, and the WFC3 IR Ryan et al. (2005), and the red ACS GRAPES and detector). PEARS grism observations of Pirzkal et al. (2005, 2009), respectively, each of which constrained the 6. SCIENCE RESULTS FROM THE ERS scale-height of Galactic L & T dwarfs to about 300– MOSAICS 400 pc. 6.1. The Panchromatic ERS Star Counts to In the bluest three WFC3 filters F225W, F275W, < AB ∼ 26 mag and F336W, the WFC3 ERS finds only 22, 18, and Fig. 11a shows the star counts and galaxy num- 27 faint stellar candidates, respectively. Most of ber counts in the 10-band ERS images. The star- these stars have 10-band colors red enough to in- > dicative of Galactic halo K–M type stars and possi- galaxy separation breaks down for AB ∼ 26 mag in > bly a few L & T dwarfs, but a few of the faint ERS BVizYJH. However, for AB ∼ 22 mag galaxies out- number stars by a large margin, so the more un- stars are blue. The latter could be faint Galactic white dwarfs. The effective ERS UV stellar detec- certain star-galaxy separation for AB > 26 mag will ∼ tion limit of AB < 25 mag could trace a T=2×104K not affect the quality of the galaxy counts. The star ∼ counts, of course, become correspondingly harder to white dwarf with an absolute magnitude in the > range Mbol'+15 to +10 mag (Harris et al. 2006) do for AB ∼ 25 mag, and so additional criteria will need to be invoked to identify faint very blue or very out to a distance of 1–10 kpc, respectively, assum- red stars with high confidence. ing modest bolometric corrections. The stellar ERS The solid black line in the F850LP panel of Fig. objects are obviously of interest by themselves, and < will be studied in more detail in a future ERS paper. 10 are the spectroscopic star counts to AB ∼ 25 mag from the HST GRAPES and PEARS ACS grism It suffices to say here that the ERS data also has surveys of Pirzkal et al. (2005, 2009). The GRAPES the potential to find faint white dwarf stars using and PEARS ACS grism surveys were done over an the WFC3 UV filters, and faint K& stars and L& T area almost twice as large as the ERS — covering dwarf through its ACS optical and WFC3 IR filters. both GOODS-N and GOOD-S — and so it is encour- 6.2. The Panchromatic ERS Galaxy Counts to aging that the Pirzkal grism star counts have the < same slope as our direct-imaging ERS star counts AB ∼ 26–27 mag in F850LP. (The slightly different star count ampli- Fig. 11c shows the ERS Galaxy count slope versus tudes are likely due to the fact that the different observed wavelength in the flux ranges AB'19–25 Galactic latitude of the GOODS-N field was aver- mag for the 3 UV WFC3 filters, AB'18–26 mag aged into the PEARS grism star counts as well). for the GOODS/ACS BViz filters, and AB'17–25 The good agreement in star counts between the two mag for the 3 WFC3 IR filters, respectively. The different survey methods confirms that — as far as GALEX FUV and NUV points of Xu et al. (2005) faint stars are concerned — our star-galaxy separa- are also plotted at 153 and 231 nm, and cover tion is accurate to at least AB'25 mag. AB=17–23 mag, again using primary color crite- Fig. 11b shows the ERS star count slope versus ria to accomplish star-galaxy separation at the 4–6” observed wavelength in the flux ranges AB'19–25.5 FWHM GALEX resolution. mag for the 3 UV WFC3 filters, AB'16–26 mag for The galaxy counts show the well known trend of the GOODS/ACS BViz filters, and AB'15–25 mag a steepening of the best-fit power-law slope at the for the 3 WFC3 IR filters, respectively. The star bluer wavelengths, which is caused by a combina- count slope in the three WFC3 ERS UV filters are tion of the more significant K-correction and the poorly determined because of the paucity of stars shape of the galaxy redshift distribution at the se- seen overall in the UV. The GALEX FUV and NUV lection wavelength. The galaxy count slope changes points of Xu et al. (2005) are also plotted at 153 and significantly from a'0.44 dex/mag at 0.23 µm wave- 231 nm, and cover AB=17–23 mag, using primary length to a'0.26 dex/mag at 1.55 µm wavelength. color criteria to accomplish star-galaxy separation Hence, the galaxy counts have a slope flatter than at the 4–6” FWHM GALEX resolution. the Euclidean value of a'0.6 dex/mag, but below The faint-end of the Galactic star count slope is the Balmer break the UV galaxy counts are steeper remarkably flat at all wavelengths from the Balmer than the a ≤0.4 dex/mag value required for their break to the near-IR, although it is rather poorly de- sky-brightness integral to converge, if they were to > termined from the ERS data alone below the Balmer continue with this power-law slope for AB ∼ 27 mag. 12

That is, the UV galaxy counts will have to turn over Hence, in principle, the counts should comparable > with a slope flatter than ≤0.4 dex/mag for AB ∼ 27 over the entire flux range AB'10–30 mag, except for mag. In §6.3 we confirm that they indeed do so. the effects of cosmic variance. The latter is small for the panchromatic ∼116 deg2 GAMA survey, but it 6.3. The Panchromatic Galaxy Counts for could be important for the other panchromatic sur- < < vey areas plotted, such as the ERS and the HUDF 10 ∼ AB ∼ 30 mag which are both (disjoint) parts of GOODS-S, as well Fig. 12a–12j show the galaxy number counts in as the HDF-N (in GOODS-N) plus the HDF-S and the 10-band ERS images compared to a number of various other ground-based surveys in the U-band. other panchromatic surveys at the bright end and Where the panchromatic GAMA and HUDF at the faint end. At the bright end, we added counts in Fig. 12a–12j overlap with the panchro- the GAMA survey (Driver et al. 2009) counts in matic ERS counts, the agreement is in general quite NUV+uvgriYJH, which cover AB=10–21 mag (Xu good, except at the bright-end of the ERS, which et al. 2005; Hill et al. 2010a). GAMA also uses suffers from cosmic variance, and from the fact that matched aperture Kron magnitudes, but GAMAs the GOODS/ERS field was chosen to be devoid of star-galaxy separation is quite different than for the objects much brighter than AB=18 mag. As a con- ERS, since there is a significant potential for stellar sequence, the bright end of the ERS galaxy counts contamination due to stars dominating the number suffers somewhat from incompleteness. The same is counts at brighter magnitudes. The GAMA sepa- true for the bright end of the HUDF, which suffers ration makes use of rPSF –rP etro to determine how similarly from cosmic variance and the avoidance of extended a source is, and the strong stellar locus on objects much brighter than AB∼21–22 mag when the (g–i) vs. (J–K) color-color diagram. A detailed that field was selected for the ultradeep HST sur- discussion on the star galaxy separation methods vey. The HUDF is adjacent to, but does not over- used in the GAMA survey is given by Baldry et al. lap with the ERS area (see Fig. 3). The BVizYJH (2010). HUDF galaxy counts are in good agreement with At the faint end, we added the HUDF counts the ERS galaxy counts to the flux levels where the in BViz from Beckwith et al. (2006), which cover panchromatic ERS counts are considered complete AB=24–30 mag, and the HUDF counts from the < (AB ∼ 26–27 mag in Fig. 11a). Bouwens et al. (2009) YJH data as compiled by Yan The good agreement between these various sur- et al. (2009), which cover AB=24–30 mag. In the veys also implies that the flux scales of the panchro- UV, we added WHT U-band counts and the HDF- matic ERS counts are approximately correct. Had North and South F300W counts as compiled by one of the ERS filter zeropoints been off signifi- Metcalfe et al. (2001), the deep LBT U-band counts > cantly ( ∼ 0.1–0.2 mag), we would have noticed this of Grazian et al. (2009), as well as the deep HST as a significant offset between the counts. Note STIS counts of Gardner et al. (2000). > this argument only holds for the counts at AB ∼ 20 We note that the filter systems in the compar- mag, as these cover the same general area of sky in isons in each of Fig. 12a-12j are not identical, al- the GOODS-S field, so that cosmic variance effects though they were chosen if every case to be as close likely affect these counts similarly over such scales as possible in wavelength (typically to within 0.1 ( < 30%, Somerville et al. 2004). dex). Hence where necessary, small but appropri- ∼ ate AB-mag offsets or color transformations were 6.4. Modeling the Panchromatic Galaxy Counts made between the filters and models used in each < < plot, following Metcalfe et al. (2001) or Windhorst for 10 ∼ AB ∼ 30 mag et al. (1991). These corrections generally contain Fig. 12a–12j over-plots simple galaxy evolution some color dependence. When applying them as a models described by Driver et al. (1998). In sum- single AB-flux scale correction for a given survey mary, these contain the best available local panchro- can introduce uncertainties of order 0.1-0.2 mag in matic LFs in ugrizYJHK from the MGC survey the flux-scale used, which over the entire AB=10–30 by Hill et al. (2010b), and in the UV using the mag range shown in Fig. 12a–12j is not noticeable. GALEX LFs of Robotham & Driver (2010). These However, for a detailed set of modeling in subse- LFs will be updated in future work with GAMA quent work, more subtle flux and color dependent redshifts from the 3 year AAT survey that was corrections may need to be applied. concluded in May 2010. This GAMA data set is < The error bars in each of the counts are Pois- 99% redshift-complete for rP etro ∼ 19.4 mag, and son, and therefore do not include effects from cosmic will provide high-fidelity, matched-volume panchro- variance. To increase the statistics where necessary matic LFs, which will be essential to interpret the at the bright-end of each survey, bins were combined higher redshift ERS work. Robotham et al. (2010) logarithmically using the local slope of the counts, give further details on the GAMA redshift survey. following Windhorst et al. (1984). At the faint-end, Based on these currently available, best panchro- the counts are not plotted in Fig. 12a–12j beyond matic local LF’s, Fig. 12a–12j show from AB=10– the level where these are deemed complete. Based 30 mag either a pure luminosity evolution model on Fig. 9 and 11, this occurs in general where the (PLE) with e(z) = 2.5 log10[(1+z)β] with PLE ex- counts turn over significantly within the error bars. ponent β=1 (dotted lines), or a number density evo- 13

γ lution (NDE) model with n(z) = no [(1+z) ] with Ryan et al. (2010) present ERS samples of faint UV- exponent γ=1 (short-dashed lines, or a combina- dropouts and very high redshift red galaxies, respec- tion of both models indicated by both β=γ=1 (long- tively. Many more such studies can be done with the dashed lines). The solid lines show in all cases the ERS data. non-evolving model (β=γ=0). These models are de- Fig. 14 shows panchromatic postage stamps of ob- scribed in more detail by Driver et al. (1995) and jects with interesting morphological structure in the Cohen et al. (2003). 10-band ERS color images of the CDF-S, yielding Over the entire flux range AB=10-30 mag, Fig. high signal-to-noise ratio detections of galaxies re- 12a–12j show that no single PLE model fits all sembling the main cosmological parameters H0 , the available galaxy count data, since there is not Ω, ρo, w, and Λ, respectively. The panchromatic enough volume at high redshift to make any PLE galaxy morphology and structure in the ERS im- model fit the high observed counts, especially in ages is so rich, that a patient investigator can find the visual to near-IR (fig. 12d–12j). This means galaxies of nearly any appearance in the epoch z'1– that one has to invoke number density evolution as 3. This is because galaxy merging was in full swing > well, i.e., β ∼ 1 and some modest amount of luminos- in this epoch. > ity evolution for γ ∼ 0. However, none of the sim- ple PLE+NDE models fits the panchromatic galaxy 7. SUMMARY AND CONCLUSIONS counts either in all 10 bands simultaneously over In this paper, we presented the new HST WFC3 the entire observed range AB=10–30 mag, not even Early Release Science (ERS) observations in the with the best available panchromatic local LF from GOODS-South field. We presented the scientific ra- GAMA (Hill et al. 2010). tionale of this ERS survey, the data taking plus data In particular, Fig. 12d–12j show that at the reduction procedures of the panchromatic 10-band faint-end of the galaxy counts in YJH there ex- ERS mosaics. We described in detail the procedure ists a significant excess in object numbers. These of generating object catalogs across the 10 different could be either an additional population at high ERS filters. redshifts, and/or — more likely — lower luminosity The new WFC3 ERS data provide calibrated, objects at lower redshifts. A detailed discussion of drizzled mosaics in the mid-UV filters F225W, these possibilities is beyond the scope of the cur- F275W, and F336W, as well as in the near-IR fil- rent paper, but the reader is referred to Bouwens ters F098W (Ys), F125W (J), and F160W (H) in et al. (2009) and Yan et al. (2009) for a discussion 1–2 HST orbits per filter. Together with the ex- of some of these possibilities. More realistic mod- isting HST Advanced Camera for Surveys (ACS) els would be ones in which the faint-end Schechter GOODS-South mosaics in the BVi’z’ filters, these LF slope α also evolves with redshift, as has been panchromatic 10-band ERS data cover 40–50 square suggested by e.g., Ryan et al. (2007) and Khochfar arcmin at 000.07–000.15 FWHM resolution and 000.090 < et al. (2007). In particular, the faint-end slope α has multidrizzled pixels to depths of AB ∼ 26.0–27.0 mag < too steepen significantly at higher redshifts, which (10-sigma) for point sources and AB ∼ 25.5–26.5 mag was suggested by e.g., Yan & Windhorst (2004), for compact galaxies. Ryan et al. (2007) and Hathi et al. (2010). This We also described the high quality star-galaxy will be subject of a future, more detailed model fit- separation made possible by the superb resolution ting paper of the panchromatic galaxy counts. For of HST/WFC3 and ACS over a factor of 10 in wave- < now, we conclude that any more sophisticated model length to AB ∼ 25–26 mag from the UV to the near- of galaxy assembly — including hierarchical merg- IR, respectively, as well as the reliability and com- ing and or downsizing with or without feedback — pleteness of the object catalogs from the ERS mo- will need to at least reproduce the overall constraints saics. Our main scientific results are: provided by the current panchromatic galaxy counts < < 1) We present the resulting Galactic star counts and from 10 ∼ AB ∼ 30 mag over a full factor of 10 in galaxy counts in 10 different filters. From the ERS wavelength. data, these could be accurately determined from < AB'19 mag to AB ∼ 26 mag over a full factor of 6.5. Interesting Classes of Objects in the 10 in wavelength from the mid-UV to the near-IR. Panchromatic ERS mosaics 2) Both the Galactic stars counts and the galaxy Fig. 13 shows several panchromatic postage stamps counts show mild but significant trends of decreas- of lower redshift early-type galaxies in the ERS with ing count slopes from the mid–UV to the near-IR nuclear star-forming rings, bars, or other interest- over a factor of 10 in wavelength. ing nuclear structure. Each postage stamp is dis- The faint-end of the Galactic star count slope is played at a slightly different color stretch, that best remarkably flat at all wavelengths from the Balmer brings out the UV nuclear structure. (Rutkowski break to the near-IR. The values of the best fit et al. (2010) present examples of ellipticals at z'0.3– power-law star count slope is in general of order 1 with various amounts of UV-excess flux. These 0.03–0.05 dex/mag above the Balmer break, i.e. galaxies are red and relatively featureless in the op- well below Euclidean value of 0.6 dex/mag. This tical, but show considerable emission in the UV, shows that to a depth of AB'25–26 mag at all wave- which is not all point-like. Hathi et al. (2010) and lengths in the range 0.2–2 µm, the ERS images are 14 quickly looking through most of the thick stellar star-forming rings and bars, and/or weak AGN ac- disk of our Galaxy at intermediate latitudes. tivity, and some objects of other unusual appear- The galaxy count slope changes significantly from ance. a'0.44 dex/mag at 0.23 µm wavelength to a'0.26 The panchromatic ERS data base is very rich in dex/mag at 1.55 µm wavelength, showing the structural information at all rest-frame wavelengths well known trend of a steepening of the best-fit where young or older stars shine during the peak power-law slope at the bluer wavelengths. This is epoch in the cosmic star-formation rate (z'1–2), caused by a combination of the more significant K- and constitutes a unique new HST data base for correction and the shape of the galaxy redshift dis- the community to explore in the future. tribution at the selection wavelength. 3) We combine the 10-band ERS counts with This paper is based on Early Release Science ob- the panchromatic GAMA counts at the bright servations made by the WFC3 Scientific Oversight < < Committee. We are grateful to the Director of end (10 ∼ AB ∼ 20 mag), and with the ultradeep BVizYJH HUDF counts and other available HST the Space Telescope Science Institute, Dr. Matt < < Mountain, for generously awarding Director’s Dis- UV counts at the faint end (24 ∼ AB ∼ 30 mag). The galaxy counts are now well measured over the entire cretionary time for this program. We thank Drs. < < Neill Reid, Ken Sembach, Ms. Tricia Royle, and flux range 10 ∼ AB ∼ 30 mag over nearly a of 10 in wavelength. the STScI OPUS staff for making it possible to have the ERS data optimally scheduled. We also 4) We fit simple galaxy evolution models to these thank WFC3 IPT and the GSFC WFC3 Project panchromatic galaxy counts over this entire flux for their very hard dedicated work since 1998 to < < range 10 ∼ AB ∼ 30 mag, using the best available 10- make this wonderful instrument work, and for their band local galaxy luminosity functions (LFs) from timely delivery of the essential hardware and soft- the GAMA survey (Hill et al.2010a), as well as sim- ware to make the acquisition and reduction of the ple prescriptions of luminosity and/or density evolu- new WFC3 data possible. We also thank the entire tion. While these models can explain each of the 10- GOODS team, and in particular Drs. Norman Gro- < < band counts from 10 ∼ AB ∼ 30 mag well, no single gin and Mauro Giavalisco for making the exquisite one of these simple PLE+NDE models can explain GOODS v2.0 mosaics available. the counts over this entire flux range in all 10 fil- Support for HST program 11359 was provided ters simultaneously. Any more sophisticated model by NASA through grants GO-11359.0*.A from the of galaxy assembly, including hierarchical merging Space Telescope Science Institute, which is oper- and or downsizing with or without feedback, will ated by the Association of Universities for Re- need to at least reproduce the overall constraints search in Astronomy, Inc., under NASA contract provided by the current panchromatic galaxy counts NAS 5-26555. RAW also acknowledges support < < from 10 ∼ AB ∼ 30 mag over a full factor of 10 in from NASA JWST Interdisciplinary Scientist grant wavelength. NAG5-12460 from GSFC. HY is supported by the These surveys also emphasize the need for ex- long-term fellowship program of the Center for Cos- > tending this work to AB ∼ 30 mag and beyond at mology and AstroParticle Physics (CCAPP) at The longer wavelengths, which the James Webb Space Ohio State University. Telescope will do for wavelengths in the range 1–28 We thank the STS-125 astronauts for risking their µm after its launch in 2014. The unique abilities lives during the Shuttle Servicing Missions SM4 to of WFC3 to do deep UV imaging will need to be Hubble, and for successfully installing WFC3 into explored fully before that time. HST in May 2009. We dedicate this paper to the memory of the STS-107 Columbia Shuttle astro- 5) We show examples of interesting panchromatic nauts and of Dr. Rodger Doxsey, who during their faint galaxy structure in intermediate redshift ob- lives contributed so much to the Space Shuttle and jects, including early-type galaxies with nuclear the Hubble Space Telescope projects.

REFERENCES Abraham, R. G.; Ellis, R. S., Fabian, A. C., Tanvir, N. R., Cohen, S. H., et al. 2006, ApJ, 639, 731 & Glazebrook, K. 1999, MNRAS, 303, 641 Cohen, S. H., et al. 2010, in preparation Abraham, R. et al. 2007, ApJ, 669, 184 Conselice, C. J. Bershady, M. A., Dickinson, M., & Afonso, J., Mobasher, B., Koekemoer, A., Norris, R. P., & Papovich, C. 2003, AJ, 126, 1183 Cram, L. 2006, AJ, 131, 1216 Crockett, R. M., et al. 2010, ApJ, submitted Baldry, I. K. et al. 2010, MNRAS, 404, 86 De Lucia, G., Springel, V., White, S. D. M., Croton, D., & Beckwith, S. V., Stiavelli, M., Koekemoer, A. M., et al. Kauffmann, G. 2006, MNRAS, 366, 499 2006, AJ, 132, 1729 Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, Bertin, E. & Arnouts, S. 1996, A&AS, 117, 393 433, 604 Bond, H. E., & Kim Quijano, J., et al. 2007, ”Wide Field Dickinson, M., Papovich, C., Ferguson, H. C., & Budavri, Camera 3 Instrument Handbook, Version 1.0” T. 2003, ApJ, 587, 25 (Baltimore: STScI) Dickinson, M., et al. 2004, ApJ, 600, L99 Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000 Driver, S. P., Windhorst, R. A., Ostrander, E. J., Keel, W. Chandar, R., et al. 2010, ApJ, submitted C., Griffiths, R. E., & Ratnatunga, K. U. 1995, ApJL, Cohen, S. H., Windhorst, R. A., Odewahn, S. C., 449, L23 Chiarenza, C. A. T., & Driver, S. P. 2003, AJ, 125, 1762 15

Driver, S. P., Fernandez-Soto, A., Couch, W. J., Odewahn, Nonino, M., Dickinson, M., Rosati, P. et al. 2009, ApJS, S. C., Windhorst, R. A., Phillipps, S., Lanzetta, K., & 183, 244 (0906.4250v1) Yahil, A. 1998, ApJL, 496, L93 Petrosian, V. 1976, ApJ, 209, L1 Driver, S. P., et al. 2009, Astron. & Geophys. 50, 5.12 Pirzkal, N., et al. 2004, ApJS, 154, 501 Dopita, M. A., et al. 2010, ApJ, 710, 964 Pirzkal, N., et al. 2005, ApJ, 622, 319 Ferreras, I., et al. 2009, ApJ, 706, 158 Pirzkal, N., et al. 2009, ApJ, 695, 1591 Gardner, J. P., Satyapal, S. 2000, AJ, 119, 2589 Rhoads, J. E., et al. 2009, ApJ, 697, 942 Gardner, J. P., Brown, T. M., Ferguson, H. C. 2000, ApJL, Robotham, A., et al. 2010, PASA, 27, 76 542, L79 Robotham, A., & Driver, S. P. 2010, MNRAS, submitted Giacconi, R., et al. 2002, ApJS, 139, 369 Ryan, R. E., Jr., Hathi, N. P., Cohen, S. H., & Windhorst, Giavalisco, M., Ferguson, H. C., Koekemoer, A. M., R. A. 2005, ApJL, 631, L159 Dickinson, M. et al. 2004, ApJ, 600, L93 Ryan, R. E., Jr., et al. 2007, ApJ, 668, 839 Grazian, A., Menci, N., Giallongo, E. et al. 2009, A& A, Ryan, R. E. Jr., Cohen, S. H., Windhorst, R. A., & Silk, J. 505, 1041 2008, ApJ, 678, 751 Grogin, N., et al. 2009, GOODS v2.0 mosaics, priv. comm. Ryan, R. E., Jr., et al. 2010, ApJ, submitted Harris, H. C., Munn, J. A., Kilic, M., et al. 2006, AJ, 131, Scannapieco, E., Silk, J., & Bouwens, R. 2005, ApJL, 635, 571 L13 Hathi, N. P., et al. 2008, AJ, 135, 156 Silk, J., & Rees, M. J., 1998, A&A, 331, L1 Hathi, N. P., et al. 2010, ApJ, submitted (arXiv:1004.5141) Somerville, R. S., 2004, ApJL, 600, L171 Hill, D. T., Driver S. P., Cameron E., Cross N. J. G., Liske Steinbring, E., et al. 2004, ApJS 155, 15 J., & Robotham, A. 2010a, MNRAS, 404, 1215 Straughn, A. N., et al. 2008, AJ, 135, 1624 Hill, D. T., et al. 2010b, MNRAS, submitted Straughn, A. N., et al. 2009, AJ, 138, 1022 (http://star-www.st-and.ac.uk/dth4/GAMA2v5.pdf)˜ Straughn, A. N., et al. 2010, ApJ, submitted Hopkins, P. et al. 2010, ApJ, 715, 202 Vanzella, E. et al. 2005, A&A, 434, 53 Hopkins, P., Hernquist, L., Cox, T., di Matteo, T., Windhorst, R. A., van Heerde, G. M., & Katgert, P. 1984, Robertson, B., & Springel, V. ApJS, 163, 1 A&AS, 58, 1 Kellermann, K. I., Fomalont, E. B., Mainieri, V., Padovani, Windhorst, R. A., et al. 1991, ApJ, 380, 362 P., Rosati, P., Shaver, P., Tozzi, P., & Miller, N. 2008, Windhorst, R. A., et al. 2002, ApJS, 143, 113 ApJS, 179, 71 Windhorst, R. A., Hathi, N. P., Cohen, S. H., Jansen, R. A. Koekemoer, A. M., Fruchter, A. S., Hook, R. N., & Hack, Kawata, D., Driver, S. P., & Gibson, B. 2008, J. Adv. W. 2002, in “The HST Calibration Workshop: Hubble Space Res., 41, 1965 (astro-ph/0703171) after the Installation of the ACS and the NICMOS Wong, M. H., Pavlovsky, C., & Long, K. et al. 2010, Wide Cooling System”, p. 337, Eds. S. Arribas, A. Koekemoer, Field Camera 3 Instrument Handbook, Version 2.0 & B. Whitmore. (Baltimore: STScI) (Baltimore, STScI) Komatsu, E., et al. 2010, arXiv, 1001.4538 Xu, K. et al. 2005, ApJ, 619, L11 Kron, R. G. 1980, ApJS, 43, 305 Yan, H., Dickinson, M., Eisenhardt, P. R. M., Ferguson, H. Labb´e,I., et al. 2003, AJ, 125, 1107 C.; Grogin, N. A., 2004, ApJ, 616, 63 Law, D. R., et al. 2007, ApJ, 656, 1 Yan, H., Dickinson, M., Stern, D., Eisenhardt, P. R. M. Law, D. R., et al. 2009, ApJ, 697, 2057 et al. 2005, ApJ, 634, 109 Lilly, S., J., et al. 1998, 1998, ApJ, 500, 75 Yan, H., Windhorst, R., Hathi, N., Cohen, S., Ryan, R., Maybhate, A., et al. 2010, ”ACS Instrument Handbook”, O’Connell, R., McCarthy, P. 2010, ApJ, resubmitted Version 9.0 (Baltimore: STScI) (astro-ph/0910.0077) Malhotra, S., et al. 2005, ApJ, 626, 666 Yan, H., et al. 2010, ApJ, submitted Melbourne, J., et al. 2005, ApJ 625, L27

8. APPENDIX A. THE WAVELENGTH-DEPENDENT UVIS GEOMETRICAL DISTORTION CORRECTION Given the issues related to the GDC discussed in §4.2.2, it was apparent that some astrometric residuals < remained — at the level of ∼ 2–3 multidrizzled pixels — that were a function of position across both UVIS detectors, after the shifts and rotations had been removed. A worst case is shown in the lower left panel of Fig. 4a. These residuals were apparent between the three UVIS filters, as well as relative to the GOODS astrometric frame. At the time of the data reduction, the best available Instrument Distortion Coefficients (IDCtab’s; file t982101i uv idc.fits) — which are used for the GDC — had been created using the WFC3 SMOV F 606W data alone (see URL29). This GDC did not yet include any wavelength-dependent distortion terms, though these will be measured later in Cycle 17 or beyond (see URL30), and distributed to the community by the WFC3 instrument team. Since our WFC3/UVIS data is all taken at wavelengths below 3500 A,˚ it is likely that the remaining astrometric residuals are indicative of the absence of these color- dependent terms in the currently existing GDC tables. The wavelength-dependent terms of the GDC come from the different index of refraction of the glass optics and filters in the UVIS channel of WFC3. This is in addition to the first–third order geometrical distortion coefficients, which results from the CCD tilt (which was necessary to minimize internal reflections), and from other reflective optics, and are therefore largely wavelength-independent. The ultimate goal of the UVIS part of the WFC3/ERS is to create images in F225W, F275W, and F336W filters at the same pixel scale (000.030 per pixel) and astrometric grid as the GOODS ACS images (Giavalisco et al. 2004) of the same portion of the Chandra Deep Field South. As discussed in §2, the eight ERS UVIS pointings were designed to overlap with each other, in order to achieve the best possible astrometric precision. After the extensive astrometric work described above, it turned out that we had reached the limit of what was possible with the present wavelength-independent geometrical distortion solution. We were able to align all UVIS images such that the average object was well aligned with the GOODS B-band image. However, objects near the individual UVIS field edges were not well aligned (lower left panel of Fig. 4a). Fig. B.1 on page 273 of the WFC3 Instrument Handbook (Wong et al. 2010) also suggests that in these locations

29 http://www.stsci.edu/hst/wfc3/idctab lbn 30 http://www.stsci.edu/hst/wfc3/STAN 09 2009 16 the geometrical distortion is the largest. This was most noticeable on the few stars in the overlap regions between pointings, as Fig. 4a demonstrated. Some of these stars also showed significant proper motion between the original 2004–2007 GOODS obser- vations and the 2009 ERS observations, complicating their use as astrometric fiducials. This is quite visible as large color offsets in some of the ERS bright star images in the very high resolution versions of Fig. 5a–5b, available on the URL quoted in the Figure caption). In attempting to create a 000.030 per pixel UVIS mosaic, the stars in the overlap regions of the 8 UVIS pointings were somewhat elongated (Fig. 4a), even though the exposures from the same UVIS orbit were well aligned — i.e., having round stars and astrometric residuals to a fraction of a pixel — and even though each pointing was well matched to the GOODS frame. In other words, even after correcting the WCS in the FITS headers for both the relative and the absolute position deviations, there still remained some astrometric residuals primarily at the field edges, that we suspect are likely due to the as yet unmeasured wavelength dependence of the GDC (Fig. 4b). Fig. 4b shows the astrometric residuals for four of our ERS pointings in the F336W filter (visits 25, 26, 29, 30). Each frame within a visit was registered using the multidrizzle scripts of Koekemoer (2002), which CR-cleaned and cross-correlated the images, to revise the FLT image WCS header keywords, as described in §4.2.1. These images were then drizzled with PIXFRAC=1.0 to a scale of 000.030 per pixel and to a position angle of 0 degrees (final rot=0). The PA(V3) was 111 degrees. Catalogs were then created with SExtractor (Bertin & Arnouts 1996), which were in turn matched in equatorial coordinates to a similar catalog made from the GOODS-South B-band image (GOODS v2.0; Giavalisco et al. 2004; Grogin 2009, priv. comm.). Note that all four visits in Fig. 4b show similar bimodal residuals, suggesting that this is a systematic error. We suspect that this is due to the wavelength-dependent geometrical distortion in the UV, since the only currently available distortion solution was measured in F606W. Since two distinct blobs of points occur in similar locations in all four panels of Fig. 4b, the multidrizzle images of objects seen in only one pointing — which includes 80–90% of the total ERS area (see Fig. 3 and 8a) — are round at 000.090 pixel sampling. However, the images of the brighter objects in the overlap areas between mosaic pointings — 10–20% of the total area — are not always completely round, as shown in Fig. 4a. When Multi-drizzling the four exposures, each individual object is well aligned with itself. However, in the WFC3 field corners, an object can be offset from the GOODS WCS by ±000.1–000.2 (Fig. 4a–4b). In the area of field overlap, an object in the ERS image corners or at a visit’s field edges can thus be misaligned with itself by ∼000.1–000.2, and therefore be elongated (see Fig. 4a), but still be on average on top of the GOODS v2.0 WCS. In other words, with the current wavelength-independent GDC and multidrizzle, we cannot have it both ways for a given object in the image borders of the ERS mosaic, until the wavelength-dependent GDC has been fully measured in the UV, and a sufficient number of dither-points is available to take full advantage of the full 000.0.0395 pixel sampling of the WFC3 UVIS images. This is at the moment not the case. Therefore, in this paper we use the total-flux measurements from the UVIS images that were multidrizzled at 000.090/pixel, and use large enough apertures that the total flux measurements are reliable. 17

Table 1. Filters, Exposure Times, & Depths of the WFC3 ERS & GOODS-S ACS data. ======Channel GRISM GRISM Filter1 Filter2 Filter3 Filter4 TOTAL ORBITS ------WFC3/UVIS G280 F225W F275W F336W Orbits -- 2 2 1 5x8 = 40 Direct Depth (AB) -- 26.3 26.4 26.1 nJy -- 110 100 132 ACS/WFC G800L F435W F606W F775W F850LP Orbits PEARS 3 2.5 2.5 5 8x20=160 GOODS Depth (AB) 27.0 28.5 28.5 27.8 27.3 nJy 58 14 14 27 44 WFC3/IR G102 G141 F098M F125W F160W Orbits 2 2 2 2 2 6x10= 60 Direct Depth (AB) 25.2 25.5 27.2 27.5 27.2 1x 4= 4 Grism nJy 230 230 48 36 48 Sub = 64

------TOTAL WFC3 ERS ORBITS (Aug.--Oct. 2009) 104 TOTAL ACS GOODS ORBITS inside ERS (2003--2007) 160 TOTAL ERS ORBITS (2003--2009) 264 ------

Notes to Table 1: The listed depth is the 50% completeness for 5-sigma detections in total SExtractor AB-magnitudes for typical compact objects at this flux level (circular aperture with 000.4 radius; 0.50 arcsec2 aperture), as derived from Fig. 9. For spectral continuum detection in the G102 and G141 grisms, these limits are about 2.0–1.8 mag brighter, respectively. We note that the pre-flight WFC3 ETC sensitivity values were up to 0.15–0.3 mag more conservative in both the UVIS and the IR than the in-flight values observed here in Table 1a and Fig. 9, as also shown in Fig. 2. 18

Table 2. ERS Filters, PSFs, Zero-points, Sky-background, and Effective Area ======HST- ERS Central Filter PSF Zeropoint Sky-back Effective Instrument Filter Lambda FWHM FWHM (AB-mag@ ground Area /Mode (mum) (mum) (") 1e-/sec) mag/(")^2 (arcm^2) ------WFC3/UVIS F225W 0.2341 0.0547 0.092 24.06 25.46 53.2 WFC3/UVIS F275W 0.2715 0.0481 0.087 24.14 25.64 55.3 WFC3/UVIS F336W 0.3361 0.0554 0.080 24.64 24.82 51.6 ACS/WFC F435W 0.4297 0.1038 0.080 25.673 23.66 72.4 ACS/WFC F606W 0.5907 0.2342 0.074 26.486 22.86 79.2 ACS/WFC F775W 0.7764 0.1528 0.077 25.654 22.64 79.3 ACS/WFC F850LP 0.9445 0.1229 0.088 24.862 22.58 80.3 WFC3/IR F098M 0.9829 0.1695 0.129 25.68 22.61 44.8 WFC3/IR F125W 1.2459 0.3015 0.136 26.25 22.53 44.7 WFC3/IR F160W 1.5405 0.2879 0.150 25.96 22.30 44.7 ------

Note 1: The panchromatic PSF FWHM was measured from ERS stars as in Fig. 7, and includes the contribution form the OTA and its wavefront errors, the specific instrument pixel sampling or Modulation Transfer Function (MTF). Note 2: The WFC3 zero-points are in AB magnitudes for 1.0 e− /sec from this URL31. Note 3: The GOODS BViz sky-background values are from Hathi et al. (2008). Note 4: The effective areas used in this paper are in units of arcminutes squared for the effective number of WFC3 or ACS tiles available and used. Note 5: The GOODS v2.0 BVi’z’ data release is from this URL32. Note 6: The panchromatic effective ERS area was derived from Fig. 8 for the WFC3 mosaics and the relevant GOODS tiles.

31 http://www.stsci.edu/hst/wfc3/phot zp lbn 32 http://archive.stsci.edu/pub/hlsp/goods/v2/h goods v2.0 rdm.html 19

Fig. 1a. (TOP panel) The full panchromatic HST/WFC3 and ACS filter set used in the ERS imaging of the GOODS CDF-South field. Plotted is the overall system throughput, or the HST Optical Telescope Assembly throughput OTA × filter transmission T × detector QE. (MIDDLE and BOTTOM panels) Spectral energy distributions for two model galaxies with ages of 0.1 and 1 Gyr and redshifts of z=0, 2, 4, 6, 8, are shown as black, blue, green, yellow, and red curves, respectively. These show that the UVIS filters sample the Lyman-α forest and Lyman continuum breaks, while the near-IR filters will probe the 4000 A˚ and Balmer breaks at these redshifts. Additional photometry is available from ground-based VLT K-band imaging, and Spitzer/IRAC imaging at 3.5, 4.5, 5.6 and 8.5 micron. 20

Fig. 1b. Mass sensitivity of the WFC3 ERS filters used. The three panels show the depth of the WFC3 ERS images as horizontal bars, compared to evolving galaxy models of different masses at three redshifts. 9 The top panel is for z=1 and a stellar mass of 10 M . The solid lines represent nearly instantaneous bursts, and the dashed lines are for declining star-formation with a 1 Gyr e-folding time. The colors refer to ages of 10, 100, 500, 1000, 1400, 2000, and 3000 Myrs from cyan to red. The center panel shows similar 9 10 models at z=1.5 for 4 × 10 M , while the lower panel is for z=2 and masses of 10 M . At z=1, the WFC3 ERS reaches masses of 0.02 M∗ for reasonable star-formation histories. Both the flux scale and the wavelength scale are logarithmic, illustrating WFC3’s exquisite panchromatic coverage and sensitivity in probing the stellar masses of galaxies through its IR channel, and the star-formation in galaxies through its UVIS channel. 21

Fig. 2. On average, the on-orbit WFC3 UVIS sensitivity is 6–18% higher than the predicted pre-launch sensitivity from the ground-based thermal vacuum test (2a; left panel), and the on-orbit WFC3 UVIS sensitivity is 9–18% higher (2b; right panel). 22

Fig. 3. Layout of the GOODS CDF-South field and its WFC3 ERS visits footprint. The 4×2 ERS UVIS mosaic is superposed in blue, and the 5 ×2 ERS IR mosaic is superposed in red. The UVIS fields are numbered from left to right, with UVIS fields 1–4 in the top row and UVIS fields 5–8 in the bottom row. The IR fields are numbered from left to right, with IR fields 1–5 in the top row and UVIS fields 6–10 in the bottom row. The dashed blue and green boxes indicate the location of the ACS parallels to the ERS WFC3 images. The ERS IR G102 and G141 grism field is shown by the green box, and overlaps with one of the PEARS ACS G800L grism field. The black box shows the ACS Hubble Ultra Deep Field in the CDF-South. The ERS program was designed to image the upper ∼20% of the GOODS-South field in six new WFC3 filters: F225W, F275W, and F336W in the UVIS channel, and F098M, F125W, and F160W in its IR channel. Further details are given in Tables 1–2. The exact pointings coordinates and observing parameters for all pointings in HST ERS program 11359 can be obtained from this URL33.

33 //www.stsci.edu/observing/phase2-public/11359.pro 23

Fig. 4a. Star imaged in the overlap region (lower right panel) between UVIS visit 25 (upper left panel) and visit 26 (upper right panel) in the F336W filter. While the star is properly processed by multidrizzle in these individual visits (i.e., its images are round in the upper left and upper right panels), it is clearly displaced by (000.061, 000.301) in its WCS location between these two visits. This is due to the wavelength-dependent geometrical distortion correction (GDC), which results in this star — and some other objects in the mosaic border regions — being elongated by approximately this amount when a multidrizzle is done of all available pointings, as can be seen in the lower left panel. A full wavelength-dependent GDC will make the images in the border regions round as well across the full multidrizzle mosaic. A proper measurement and continued monitoring of the wavelength-dependent GDC is scheduled for later in Cycle 17 and beyond. 24

Fig. 4b. The astrometric residuals for four of our ERS pointings in the F336W filter (visits 25, 26, 29, 30). Note that all four visits show similar bimodal residuals, suggesting that this is a systematic error. We suspect that this is due to the wavelength-dependent geometrical distortion in the UV, since the only currently available distortion solution was measured in F606W. Since two distinct blobs of points occur in similar locations in all four panels, the multidrizzle images of objects seen in only one pointing — which includes 80–90% of the total ERS area (see Fig. 3 and 8a) — are round at 000.090 pixel sampling. However, the images of the brighter objects in the overlap areas between mosaic pointings — 10–20% of the total area — are not always round, as can be seen in Fig. 4a. 25

Fig. 5a. Panchromatic 10-band color image of the entire ERS mosaic in the CDF-South. [This image is only displayed in the electronic version of this journal paper.] Shown is the common cross sections between the 4×2 ERS UVIS mosaics, the GOODS v2.0 ACS BViz, and the 5×2 ERS IR mosaics. The total image shown is 6500 × 3000 pixels, or 9.75 × 4.5 arcmin. We used color weighting of the 10 ERS filters as described in the text, and log(log) stretch. A higher resolution version of this image may be obtained from this URL34.

34 http://www.asu.edu/clas/hst/www/wfc3ers/ERS2 loglog.tif 26

Fig. 5b. Zoom of the panchromatic 10-band ERS color image in the CDF-South (see Fig. 5a). The total image is 6500 × 3000 pixels, or 9.75 × 4.5 arcmin. A higher resolution version of this image may be obtained from this URL35.

35 http://www.asu.edu/clas/hst/www/wfc3ers/ERS2 gxysv4ln.tif 27

Fig. 6a. Color reproduction of a bright but unsaturated star image in the 10-band ERS color images in the CDF-South. Fig. 6b. Same for a double star. These images give a qualitative impression of the significant dynamic range in both intensity and wavelength that is present in these panchromatic ERS images. 28

Fig. 7a (left panels). Stellar images, and Fig. 7b (right panels). stellar light-profiles in the panchromatic 10- band ERS color image in the CDF-South. Note the progression of λ/D, except that at wavelengths shorter than the V-band, HST is no longer diffraction limited, resulting in wider image-wings and a somewhat larger > fraction of the stellar flux visible outside the PSF-core. The “red halo” at λ ∼ 0.8 µm is due to noticeable Airy rings in the stellar images in the WFC3 IR channel, and the red halo in the ACS z-band (Wong et al. 2010). 29

Fig. 8a (TOP LEFT panels). Cumulative distribution of the maximum pixel area that has the specified fraction of total orbital exposure time in each ERS filter. These effective areas must be quantified in order to properly do the object counts in Fig. 11a. About 50 arcmin2 has ∼80% of the exposure time or ∼90% of the intended UV sensitivity. Fig. 8b (BOTTOM MIDDLE panels) Same as Fig. 8a, but for the GOODS v2.0 images in BViz. Fig. 8c (TOP RIGHT panels) Same as Fig. 8a, but for the ERS mosaics in the IR. About 40 arcmin2 has ∼80% of the exposure time in the ERS IR mosaics, or ∼90% of the intended IR sensitivity. The overall WFC3 UVIS – IR sensitivity is 9–18% better than predicted from the ground-based thermal vacuum tests (see § 2 and Fig. 2), and so in essence 100% of the intended exposure time was achieved over the 50 arcmin2 UVIS images and the 40 arcmin2 IR images. 30

Fig. 9. Panchromatic completeness functions of the galaxy number counts in the ERS images. Violet, dark and light blue lines indicate WFC3/UVIS F225, F275, F336W; and orange through red lines indicates WFC3/IR F098M, F125W, and F160W, respectively. These were derived by Monte Carlo insertion of faint objects into the ERS images, and plotting the object recovery ratio as a function of total magnitude). The ACS BViz samples in the ERS area (not shown here) are substantially deeper, and are substantially complete < to AB ∼ 28.5–27.5 mag, respectively (Table 1). 31

< Fig. 10. Panchromatic star-galaxy separation in the 10-filter ERS images. Objects with re ∼ 0.9 FWHM, where FWHM from Table 2 (or 000.07–000.15 FWHM) are image defects (plotted as black dots). Objects in the thin vertical filaments (plotted as red dots) immediately larger than this are stars. Objects to the right of the full-drawn slanted lines are galaxies (plotted as black dots). The star-galaxy separation breaks down at > AB ∼ 26–27 mag in UVBVizYJH, respectively. Details are discussed in the text. [ A full resolution version of this figure is available on this URL36.

36 http://www.asu.edu/clas/hst/www/wfc3ers/ERS2 gxysv4ln.tif 32

Fig. 11a. Panchromatic star counts (red asterisks) and panchromatic galaxy number counts in the ERS images (black dots), with the optimized star-galaxy separation from Fig. 10. All 10 ERS filters are shown in units of object numbers per 0.5 mag per deg2, but in the three bluest filters (F225W, F275W, and F336W) the bins were doubled to 1.0 mag in width to improve statistics. The solid black line in the F850LP panel are the spectroscopic star counts from the HST GRAPES and PEARS ACS grism surveys of Pirzkal et al. (2005, 2009), which are in good agreement with our F850LP star counts. 33

Fig. 11b. (LEFT) ERS Star count slope (filled circles) versus observed wavelength in the flux ranges AB'19– 25.5 mag for the 3 UV WFC3 filters, AB'16–26 mag for the GOODS/ACS BViz filters, and AB'15–25 mag for the 3 WFC3 IR filters, respectively. The faint-end of the Galactic star count slope is remarkably flat at all wavelengths from the mid–UV to the near-IR, with best fit power-law slopes in general of order 0.03–0.06 dex/mag. The two bluest points at 153 and 231 nm are from the GALEX star counts of Xu et al. (2005; open squares), which cover AB=17–23 mag. 11c. (RIGHT) ERS Galaxy count slope (filled circles)versus observed wavelength in the flux ranges AB'19– 25 mag for the 3 UV WFC3 filters, AB'18–26 mag for the GOODS/ACS BViz filters, and AB'17–25 mag for the 3 WFC3 IR filters, respectively. The two bluest points at 153 and 231 nm are from the GALEX galaxy counts of Xu et al. (2005; open squares), which cover AB=17–23 mag. The galaxy counts show the well known trend of a steepening of the best-fit power-law slope at the bluer wavelengths, which is caused by a combination of the more significant K-correction and the shape of the galaxy redshift distribution at the selection wavelength. 34

Fig. 12. Galaxy number counts for the entire flux range AB=10–30 mag in the F225W (12a), F275 (12b), and F336W (1c) filters. For comparison at the bright end, we added to the ERS number counts from Fig. 11 the panchromatic GAMA survey (Driver et al. 2009) counts in NUV+uvgriYJH, which cover AB=10–21 mag (Xu et al. 2005; Hill et al. 2010a). At the faint end, we added the WHT U-band and HDF-North and South F300W counts of Metcalfe et al. (2001), the deep LBT U-band counts of Grazian et al. (2009), as well as the deep HST STIS counts of Gardner et al. (2000). Best fit luminosity and number density evolution models range from AB=10–31 mag (see Driveret al. 1998; Hill et al. 2010a). 35

Fig. 12 (cont). Galaxy number counts for the entire flux range AB=10–30 mag in the F435W (12d), F606W (12e), F775W (12f), and F850LP (12g) filters. For comparison at the bright end, we added to the ERS number counts from Fig. 11 the panchromatic GAMA survey (Driver et al. 2009) counts in NUV+uvgriYJH, which cover AB=10–21 mag (Xu et al. 2005; Hill et al. 2010a). At the faint end, we added the HUDF counts in BViz from Beckwith et al. (2006), which cover AB=23–30 mag. Best fit luminosity and number density evolution models range from AB=10–31 mag (see Driveret al. 1998; Hill et al. 2010a). 36

Fig. 12 (cont). Galaxy number counts for the entire flux range AB=10–30 mag in the F098M/F105W (12h), F125W (12i) and F160W (12j) filters. For comparison at the bright end, we added to the ERS number counts from Fig. 11 the panchromatic GAMA survey (Driver et al. 2009) counts in NUV+uvgriYJH, which cover AB=10–21 mag (Xu et al. 2005; Hill et al. 2010a). At the faint end, we added the HUDF counts in YJH from the Bouwens et al. (2009) YJH data as compiled by Yan et al. (2009), which cover AB=24–30 mag. In the UV, we added WHT U-band and HDF-North and South F300W counts of Metcalfe et al. (2001), the deep LBT U-band counts of Grazian et al. (2009), as well as the deep HST STIS counts of Gardner et al. (2000). Best fit luminosity and number density evolution models range from AB=10–31 mag (see Driveret al. 1998; Hill et al. 2010a). 37

Fig. 13. Panchromatic postage stamps of early-type galaxies in the ERS with nuclear star-forming rings, bars, or other interesting nuclear structure. Each postage stamp is displayed at a slightly different color stretch that best brings out the UV nuclear structure. 38

Fig. 14. Panchromatic postage stamps of objects with interesting morphological structure in the 10-band ERS color images of the CDF-S: from left to right, high signal-to-noise ratio detections of ERS galaxies resembling the main cosmological parameters H0 , Ω, ρo, w, and Λ, respectively. These images illustrate the rich and unique morphological information available in the 10-band panchromatic ERS data set.